Abstract
The problem of numerically classifying patterns, of crucial importance in the biomedical field, is here faced by means of their fractal dimension. A new simple algorithm was developed to characterize biomedical mono-dimensional signals avoiding computationally expensive methods, generally required by the classical approach of the fractal theory. The algorithm produces a number related to the geometric behaviour of the pattern providing information on the studied phenomenon. The results are independent of the signal amplitude and exhibit a fractal measure ranging from 1 to 2 for monotonically going-forwards monodimensional curves, in accordance with theory. Accurate calibration and qualification were accomplished by analysing basic waveforms. Further studies concerned the biomedical field with special reference to gait analysis: so far, well controlled movements such as walking, going up and downstairs and running, have been investigated. Controlled conditions of the test environment guaranteed the necessary repeatability and the accuracy of the practical experiments in setting up the methodology. The algorithm showed good performance in classifying the considered simple movements in the selected sampleof normal subjects. The results obtained encourage us to use this technique for an effective on-line movement correlation with other long-term monitored variables such as blood pressure, ECG, etc.